Jump to content

  • Log In with Google      Sign In   
  • Create Account


#ActualBacterius

Posted 04 September 2012 - 07:20 AM

If I remember correctly, you must multiply the running color by the reflected contribution (component-wise) instead of adding - this can be justified via energy conservation or some spectral theory. Then you might still be off by some constant factor, but it should look more realistic. I could be wrong though - try it out. Also! This means you start your ray at values (1, 1, 1) by definition, otherwise the image will be all black, obviously.

I was expecting more attenuation on the reflections (due distance maybe?), but the reflection is allways so mirror perfect..

Yes, perfect reflection won't do this for you. If you want to add this, you need to add Fresnel reflection, which'll attenuate the reflection depending on the angle between the incident ray and the surface normal, so you'll get realistic reflections. Also, instead of perfect reflection, you can make it so the rays are mostly reflected perfectly, but slightly jittered, which will make the reflected image blur out depending on its distance to the reflection plane - but for this you need to sample multiple rays per pixel, so you'll need to make modifications for this to work (you can also sort of fake it, but I wouldn't know).

One problem is that I dont know how to let the clear/bg color with reflections on, because whenever a ray miss, it will add the clear color to the final color, making the scene too bright, so now Im returning black on ray miss...(any tips on that?)

When a ray misses you have to stop the recursion - you are done. This'll fix your problem (I think, unless I misunderstood). If your ray instantly misses without hitting anything, you can give it whatever color you want (the background color) which is only used to color those "background pixels" and never comes into play anywhere else. Ideally, your rays should always hit something (even if just a skybox/skysphere) but having a background color is fine too.

#4Bacterius

Posted 04 September 2012 - 07:12 AM

If I remember correctly, you must multiply the running color by the reflected contribution (component-wise) instead of adding - this can be justified via energy conservation or some spectral theory. Then you might still be off by some constant factor, but it should look more realistic. I could be wrong though - try it out. Also! This means you start your ray at values (1, 1, 1) by definition, otherwise the image will be all black, obviously.

I was expecting more attenuation on the reflections (due distance maybe?), but the reflection is allways so mirror perfect..

Yes, perfect reflection won't do this for you. If you want to add this, you need to add Fresnel reflection, which'll attenuate the reflection depending on the angle between the incident ray and the surface normal, so you'll get realistic reflections. Also, instead of perfect reflection, you can make it so the rays are mostly reflected perfectly, but slightly jittered, which will make the reflected image blur out depending on its distance to the reflection plane - but for this you need to sample multiple rays per pixel, so you'll need to make modifications for this to work.

#3Bacterius

Posted 04 September 2012 - 07:09 AM

If I remember correctly, you must multiply the running color by the reflected contribution (component-wise) instead of adding - this can be justified via energy conservation or some spectral theory. Then you might still be off by some constant factor, but it should look more realistic. I could be wrong though - try it out.

I was expecting more attenuation on the reflections (due distance maybe?), but the reflection is allways so mirror perfect..

Yes, perfect reflection won't do this for you. If you want to add this, you need to add Fresnel reflection, which'll attenuate the reflection depending on the angle between the incident ray and the surface normal, so you'll get realistic reflections. Also, instead of perfect reflection, you can make it so the rays are mostly reflected perfectly, but slightly jittered, which will make the reflected image blur out depending on its distance to the reflection plane - but for this you need to sample multiple rays per pixel, so you'll need to make modifications for this to work.

#2Bacterius

Posted 04 September 2012 - 07:07 AM

If I remember correctly, you must multiply the running color by the reflected contribution (component-wise) instead of adding - this can be justified via energy conservation or some spectral theory. Then you might still be off by some constant factor, but it should look more realistic. I could be wrong though - try it out.

#1Bacterius

Posted 04 September 2012 - 07:06 AM

If I remember correctly, you must multiply the running color by the reflected contribution (component-wise) instead of adding - this can be justified via energy conservation or some spectral theory. Then you might still be off by some constant factor, but it should look more realistic.

PARTNERS